Optimize performance of locked mobilizers about simbody HOT 3 CLOSED

Hmmm ... hard to see how to do this automatically. There are a few issues:

• Locks are implemented by solving for the generalized forces necessary to hold the joint at its locked value. Even when locked at position level that is not exactly the same thing as combining the bodies -- all the separate body kinematics are still available as well as the computed generalized forces (those can be necessary for deciding when to release the lock, for example).
• Locks at velocity or acceleration level use the same equations as at position level but are clearly different than combining the rigid bodies.
• At the Simbody level there is no association between force elements and mobilizers. Instead the System accumulates a set of body forces and generalized forces from all the force elements. Any given force element may contribute to any or all of those. Even if there were, it is not clear that you would want these disabled in general because the presence of a mobilizer force (however frustrated) affects the answer you get when you ask how much torque or force is being generated by the lock. So for example if you had a torsional spring acting on a pin joint, and were using a lock to emulate a torque-limited motor, the presence of the spring would affect how much torque the lock was generating and thus when it should be turned off.

I'm thinking it makes more sense to disable the force and lock the joint together in the same chunk of code, or perhaps built a custom force element that would have the desired properties. Do you have an alternate idea?

from simbody.

I'm going to close this because it doesn't seem to fit within Simbody's architecture as stated for the reasons listed above. There could be related performance tweaks possible though, please open another issue if you think of one.

from simbody.

Sounds good to me. Thanks for the update.

Kevin

On Mon, Oct 21, 2013 at 9:06 PM, Michael Sherman
[email protected]:

I'm going to close this because it doesn't seem to fit within Simbody's
architecture as stated for the reasons listed above. There could be related
performance tweaks possible though, please open another issue if you think
of one.

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/21#issuecomment-26775825
.

from simbody.